#Strategy

In one season, four major labs shipped the same idea: an agent that does real work inside your files and tools. Some of us have been working this way for months. Now it is arriving for everyone, and it changes both the tasks you repeat and the problems you solve once.

Consumption pricing is suppressing the usage the frontier AI labs need most right now. From inside the enterprise, a simple fix: make the model cheap and sell the harness by the seat, at a price a company can actually predict.

The headlines say AI is coming to replace us. Watch the people who use it best and you see the opposite. They spend less time executing and more time thinking, judging, and directing the work. AI is handing back the most human parts of the job, and to everyone, not just the people at the top.

I built a personal company brain that I talk to every night. It works because it is built for one. Scaling it to a company runs straight into governance and access control, and that is the part still worth solving.

At a recent client kickoff, employees were weighing their own tasks in dollars for the first time. Consumption pricing gives finance a clean number, but it also teaches curious people to hesitate. A look at what the meter does once it reaches the floor, and a better way to handle it.

A description of what good AI work actually feels like, why it's held up across model eras and platforms, and why it's also the shape of knowledge work going forward.

A working through-line for what changed in how we use ChatGPT, why the term "context engineering" is doing real work, and the practical moves that separate people who get good output from people who don't. Includes a 30-minute coached exercise you can run on yourself.

Anthropic and OpenAI both announced today that they're standing up implementation arms backed by billions in capital. Anthropic launched a $1.5 billion joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs. OpenAI finalized a $10 billion vehicle anchored by TPG. That confirms what operators in this space have been quietly investing in for a year: the translation layer between AI capability and business outcomes is real, it's a category, and it's not going away.

I caught myself, three hours into building a financial-modeling app, reaching for the old reflex of turning it into a public product. For the first time in twenty years I dismissed the thought on the spot. The reasoning that came up was new, and the rest of the market is arriving at the same conclusion at the same time.

Anthony Pompliano reversed his position on AI and jobs this week, and Marc Andreessen amplified it. The data Pompliano cites is real. The data on the other side, showing entry-level hiring collapse and recent-grad unemployment near a fifteen-year high, is also real. Both can be true at once because they are describing two different kinds of companies.

OpenAI launched workspace agents this week. Claude has had cloud-hosted scheduled automations for weeks. Here's what's actually different between the two, and what it means for your organization.

Anthropic's Amanda Askell says Claude slips into "criticism spirals" when users get sharp, hostile, or unpredictable. The finding is really about prompt tone as a workforce skill, and what it means for AI training programs that skip right past it.